Over the past 10 to 15 years, the focus on reliability above all has shifted towards higher efficiency, lower PUE, and lower operating costs. The shift creates an interesting paradigm with endusers split. Management wants the most efficiency, to boast green practices, and reduced costs. Their operators, agents, and actual implementers do not want their reliability to suffer. Obviously, downtime is frowned upon in our industry as the resulting loss of data may be financial or the cure for cancer.
The importance of a company’s data can not be understated — and the investment businesses make to infrastructure or in collocation is substantial. Because of this, it is imperative to understand just what happens behind the scenes at a data center and how every decision you make can directly affect the success of your business. Topics such as maintenance, staffing, the environmental controls, and disaster recovery plans are critical to understand in order to be successful, and yet the key decision maker eventually comes back to the not so simple cost of efficiency vs. the cost of reliability.
The truth of the matter is, entirely green data centers that boast efficiency over reliability are somewhat sensationalized in the industry. They match up to a customer’s sustainability mantra, and typically, the natural way in which they cool or warm the data center, or use limited hardware and technology to run the infrastructure is entirely desirable. It can be cheaper up front and the public perception of green companies is positive and therefor can be a great sales and marketing tool for customers. There is something to be said for that, but consider this: when a green data center crashes, the costs associated with that downtime may greatly exceed choosing robust infrastructure from the start.
For colocation providers — happy customers are the Holy Grail. When operations run smoothly, our business has no perceptible bumps, and our reaction to incidents great and small go undetected — we’ve done our job and our customers are happy. Redundant equipment very well might be viewed as wasteful, both as a first cost and as a long term OPEX, but it acts as a security blanket to the very real fear of having downtime for hours or days. Call anyone in New York City after Hurricane Sandy to hear firsthand accounts.
In our opinion, a much greener approach lies with proper IT and server vendor coordination. A simple pricing exercise for server RAM concluded that you can double the amount of RAM in some platforms for $2. RAM can account for upwards of 25% of server power draw, and just like redundant infrastructure, extra RAM results in lower utilization, and wasted power. There’s a lot more RAM out there than UPS systems and chillers. It might be that everyone’s looking the wrong way on the real approach to efficiency and reliability.